Dealing with “Sympathetic Bias”

Posted in Issues

Blame for the crisis in real estate lending seems to be zeroing in on the credit-rating agencies that, starting in 2000, signed off on the complex new debt instruments that have proved to be so vulnerable and opaque. As reporters Aaron Lucchetti and Serena Ng explained in The Wall Street Journal last week  (subscription required). 

“It was lenders that made the lenient loans, it was home buyers who sought out easy mortgages, and it was Wall Street underwriters that turned them into securities. But credit-rating firms also played a role in the subprime-mortgage boom that is now troubling financial markets.  Standard & Poor, Moody’s Investor Service and Fitch Ratings gave top ratings to many securities built on questionable loans, making the securities seem as safe as a Treasury bond.”

The fundamental impulse directing large new volumes of funds into real estate almost certainly arose from the collapse of the Internet bubble. Lucchetti and Ng made a good start on telling the fascinating story of how “piggyback” mortgages (one loan for a down payment and another for the balance of the sale price) and many other exotic instruments got a green light from credit-rating agencies in 2000.

Such mortgages were then widely offered, as money pulled out of the equities markets and poured into real estate lending instead. Home values, after all, were the one asset that could be trusted to continue to go up (or at least not go down). And so, for several years, up they went.  Elaborate mechanisms were devised to “securitize” the loans the banks and mortgage firms were making — underwriters would assemble them in pools as collateral to bonds, which then were sold to banks, pension funds and hedge funds seeking slightly higher returns on their investments.

Behind the scenes, the rating agencies took an active hand in devising the new securities. The Economist last week described the process this way: “Like schoolgirls asking for help with their homework, the banks would go to the agencies and ask how… the collateralized debt obligations they were putting together would score.” 

The rating firms’ compensation in turn rose with their participation in fashioning the new instruments:  they earned about twice as much evaluating a pool of mortgage loans as rating a conventional corporate bond. No wonder, then, that they were slow last year to acknowledge that their optimistic default-risk assumptions were failing to work out in practice.

“It was always about shopping around” for higher ratings,” one former Moody’s executive told the WSJ reporters.  Banks and mortgage-firm proprietors preferred to describe their efforts as seeking “best execution” or “maximizing value.”  An executive at Standard & Poor, a McGraw-Hill subsidiary, said “We don’t negotiate the criteria.  We do have discussions.”

After the corporate scandals of the Tech Wreck (Enron, Tyco International, WordCom, HealthSouth Corp and all that), the accounting firms were pilloried.  Now perhaps it is the rating agencies’ turn.  State attorneys general are investigating. Congressional hearings are being prepared. French president Nicholas Sarkozy called on G7 governments and central banks to tackle problems of financial transparency. (Portfolio magazine has a good roundup of the gathering storm.)

But there has got to be more to the story than the familiar Punch-and-Judy routines.  And, sure enough, there is.

A particularly interesting angle on the problem has been explored by Sandra Peart, of the University of Richmond, and David Levy, of George Mason University, experts on, among other things, the psychology of motive. The pair cut their teeth on the topic with a close study of the argument about equality and hierarchy that unfolded among economists and eugenicists during the nineteenth-century, described in The Vanity of the Philosopher. Recently they have been working on ethical issues that arise in statistical consulting.

They see a potential problem any time an expert is hired to give an “unbiased” opinion on a matter of interest to his client and, in a pair of recent papers, have explored it with formal models, the sort that depend heavily on Greek letters for their expression. (Sorry, neither is available on the Web yet.) In  “Sympathetic Bias,” in Statistical Methods in Medical Research, they spell out how a non-pecuniary motive that starts out merely as desire for approbation, may become affection and a willingness to fudge results when the statistical investigator is captured by the interests of the client. A more elaborate model, they note, could show how a sympathy motive could turn pecuniary as quickly as the investigator learns that “bias pays very well indeed.” 

In “Inducing Greater Transparency,” scheduled to appear in the Eastern Economic Journal in January 2008, they offer some concrete examples. They begin by noting the personal code of conduct of the heroes of statistical consulting, W. Edwards Deming, college professor, business consultant (especially to Japanese firms), and father of many modern methods of quality control. (Strictly observe professional norms, retain publication rights, prepare your own expert testimony, reserve the right to break off the engagement, etc.) Then they allow various temptations to enter their model. The right approach would take too long to explain to the client, who doesn’t understand it anyway. The client wants strong results but the data are inconclusive. The study looks pretty good except for an anomaly that the client doesn’t want to discuss.  So why not make many estimates and pick the most favorable? 

Levy and Peart’s recent work is directed towards the problem of rendering the expert witness more accountable. Indeed, the subtitle of “Inducing Greater Transparency” is “Towards the Establishment of Ethical Rules for Econometrics.” (The relatively recently-arrived profession has no code of ethics, formal or, according to many practitioners, informal.) But the problems they are examining are common to a wide range of professional experts, judges, regulators, educators, referees and watchdogs on which society depends to do its business. “Bias” and “capture” of the sort of which, most recently, the credit-rating agencies stand accused are ubiquitous. How, then, to solve what they call “the enormous problem” of harnessing small groups to serve the interests of the general public?

The minimum goal should be transparency, they say.  Other people, including the experts’ competitors, ought to be able to figure out how they get their results. But transparency won’t be forthcoming in, say, the mortgage-lending business, if rating agencies are able to pick and choose from the estimation procedures they apply to the reams of data they are given by lenders. Naturally rating firms will shade their opinions to please the lenders.  Economists call this an “incentive-compatibility” problem.

To this, Levy and Peart propose a simple solution: randomization. Rotate ratings firms the way that baseball rotates umpires. If they were assigned by lottery, rating agencies would have enhanced incentives to take the public interest into account — and diminished incentives to try to please underwriting institutions that were paying the bills.  Something of the sort was incorporated among the reforms mandated for accounting firms by the Sarbanes-Oxley Act of 2002 — the rotation every five years of the lead audit partner and the reviewing audit partner was required, for example, and more frequent changes of firms themselves was recommended. But the measures stopped well short of randomization.

Whatever the array of micro-motives, and the institutions that are designed to channel them in desirable directions, the biggest problem today may have to do with the overall architecture by which the system is conceived.  For thirty years and more, the very idea of government as a realm apart from the realm of commerce has been under probing assault. The idea that those who choose to serve as the public good must be motivated by subtly different values than those who create its private goods and services is as old as Plato. Yet the concept of countervailing powers has received relatively little attention recently, except at the level of individual professions — stock exchanges one crisis, accounting firms the next, credit rating agencies after that.

To really change the rules, it may be necessary to re-conceptualize the system from the top down, in one of those mysterious periodic swings of emphasis on public and private involvements that, at least in the United States, have been delineated by historians from Henry Adams to Arthur Schlesinger Jr. and Albert Hirschman. For the men and women who staff the credit rating agencies to behave less like business developers and more like public servants may require the sort of deep changes in preferences that are usually accomplished (or at least expressed) in an election or, rather, a series of elections.

Meanwhile, the real estate sector may not heretofore have been a standard chapter in the story of manias, panics and crashes.  Henceforth, though, it is sure to be!